Feedback

Overview

We considered the degree to which feedback ("squealing") was a problem -- another common complaint from hearing aid users[16].

There is already some influence of the device's feedback properties in the Speech Perception Benefit score, because we only considered stable settings (without audible squeals) to be appropriate. However, we additionally tested the quality of the feedback canceller in two challenging cases.

Challenge Cases

The following provocations were performed while the library scene was played in the background to ensure that the devices were amplifying:

1. Near-Ear Hand Motion: Repeatedly moving hands near the device for 10 seconds
2. Cupping: Using a hand to cup the ear 10 times in a row

We recorded from the manikin during both cases.

Blind Listening Test

Trained listeners performed blind listening tests to subjectively rate each recording on a 0--3 scale:

Inter-Rater Agreement

Three expert raters each performed two repetitions:

• Perfect agreement (same score): 77% of trials
• Strong agreement (all scores within 1 point): 100% of trials

Given this high agreement, we simply averaged the scores across raters for each challenge case.

Final Metric

The final metric was computed by summing the points across a device's 2 challenge cases and scaling to a 5-point scale.

• 5 = no audible feedback in either challenge case
• 0 = strong feedback in both cases